Amazon itself would only say in a statement that the site was "experiencing an issue." (That's one of the great PR phrases, in that you can dissect it as carefully as you want and it still communicates absolutely no information.) But Amazon also promised to deliver updates "until service has been restored," which certainly suggests that service—at that moment—was less than ideal. And, according to site traffic monitoring firm Pingdom, less than ideal accurately described Amazon Tuesday afternoon. At times, the site was taking 10 seconds to deliver a page, which is an eternity for Amazon.
Pingdom pulled Amazon's traffic reports for Tuesday, looked at its Amazon results and confirmed that the site had indeed not crashed. "They ran into performance issues that slowed down the site quite significantly, but I also verified that the site was reachable, albeit very slow," said Pingdom Web Analyst Peter Alguacil. "OK, strictly speaking, there were a handful of requests that were met with HTTP errors—both 500 internal server and 503 service unavailable errors—but nothing significant. They happened around the worst peak and are definitely a sign that their servers were temporarily overloaded for some reason."
Hosting firms are notorious for playing games with service-level agreement (SLA) wording, where they promise 99.999 percent uptime, but exclude any site slowdown short of an absolute crash for all site visitors. Should SLAs specify performance will not drop below a predetermined threshold (two seconds for a page load seems typical)?
Some of that, the hosters will argue, is under the control of the retailer. If the chain decides to fill every page with auto-play full-screen video and to then wildly market a sale without paying for additional bandwidth, it's hard to blame the server farm for the slowdown.
Either way, Amazon has for years been the most stable of any major E-tailer. When a site slowdown for a few hours makes The Wall Street Journal, that probably is a reverse compliment for your site performance.